\(L^p\)-integrability of the gradient of solutions to quasilinear systems with discontinuous coefficients. (Q2249881): Difference between revisions

The paper deals with higher integrability of weak solutions to quasilinear elliptic and parabolic systems with controlled growth on Reifenberg flat domains. The considered systems are of the type \[ D_{\alpha }(A_{ij}^{\alpha \beta }(x)D_{\beta }u^j(x)+ A_i^{\alpha }(x,u)= b_i(x,u,Du)\;\text{a.e. on }\;\Omega, \tag{1} \] \[ u(x)= 0 \;\text{ on }\;\partial \Omega \] in the elliptic case. Its parabolic counterpart reads \[ u_t^i - D_{\alpha }(A_{ij}^{\alpha \beta }(x)D_{\beta }u^j(x)+ a_i^{\alpha }(x,u)= b_i(x,u,Du)\;\text{ a.e. on }\;Q=\Omega \times (0,T),\tag{2} \] \[ u(x)= 0 \;\text{ on }\;\partial \Omega \times (0,T). \] The domain \(\Omega \) is Reifenberg \((\delta, R)\) flat, the coefficient matrix \(A\) is bounded, elliptic and \((\delta ,R)\) vanishing. In lower-order terms the coefficients \(a,b\) are Carathéodory functions and in the elliptic case satisfy the estimates \[ a_i^{\alpha }(x,u)= O(\varphi_1(x) + | u| ^{\frac {n}{n+2}}), \] \[ b_i(x,u, z)= O(\varphi_2(x) + | u| ^{\frac {n + 2}{n - 2}}), \] where \(\varphi_1 \in L^p(\Omega)\) with \(p> 2\); \(\varphi_2 \in L^q(\Omega)\) with \(q>\frac {2n}{n+2}\). The assumptions in the parabolic case are \[ | a_i^{\alpha }(x,u)| \leq \Lambda (\psi_1(x,t) + | u| ^{\frac {n+2}{n}}), \] \[ | b_i(x,u, z)| \leq \Lambda (\psi_2(x,t) + | u| ^{\frac {n + 4}{n}} + | z| ^{\frac {n+4}{n+2}}), \] where \(\psi_1 \in L^p(Q)\) with \(p> 2\); \(\psi_2 \in L^q(Q)\) with \(q>\frac {2(n+2)}{n+4}\). In the elliptic case the author proves: Theorem 2.3: Let \(u \in W^{1,2}_0(\Omega; R^N)\cap L^{\infty }(\Omega; R^N)\) be a weak solution to (1) under the above-cited conditions. Then there exists \(\delta_0>0\) such that if \(\Omega \) is \((\delta, R)\) flat and \(A_{ij}^{\alpha \beta }\) are \((\delta, R)\) vanishing with a positive \(\delta < \delta_0 <1\), then \(u \in W^{1,r}_0(\Omega ;R^N)\cap L^\infty (\Omega; R^N)\) with \(r = \min (p,q^*)\). Here \(q^*\) is Sobolev conjugate to \(q\), i.e. \(q^{*} = \frac {nq}{n-q}\) for \(q<n\) and \(q^{*} > 1\) arbitrary for \(q \geq n\). In the parabolic case the author obtains: Theorem 3.1: Let \(u \in W^{1,2}_{0,*}(Q;R^N)\cap L^{\infty }(Q; R^N)\) be a weak solution to (1) under the above-cited conditions. Then there exists \(\delta_0 > 0\) such that if \(\Omega \) is \((\delta, R)\) flat and \(A_{ij}^{\alpha \beta }\) are \((\delta, R)\) vanishing with a positive \(\delta < \delta_0 <1\), then \(u \in W^{1,r}_{0,*}(Q;R^N)\cap L^\infty (Q; R^N)\) with \(r = \min (p,q^*)\). The space \(W^{1,r}_*(Q;R^N) = L^r(0,T; W^{1,r}(\Omega; R^N))\cap W^{1,r}(0,T; W^{-1,r'}(\Omega; R^N))\) consists of those functions \(u \in W^{1,0}_r(Q; R^N)\) for which there exist functions \( g \in L^r(Q; R^N)\) and \(F \in L^r(Q; M^{N\times n})\) such that \(u_t = \text{div}F - g\) a.e. on \(Q\) in the distribution sense. Moreover, \(\| u\|_{W^{1,r}_{*}(Q;R^N)}= \| u\|_{W^{1,0}_r(Q;R^N)} + \inf \left \{\int_Q | F| ^r + | g| ^r)\right \}\). The closure of \(C_0^{\infty }(Q;R^N)\) in this norm is denoted \(W^{1,r}_{0,*}(Q;R^N)\).